[gmx-users] Problems with distance restraints

[chris.neale at utoronto.ca]

It's not really a solution, but you could try a different number of  
nodes. I have found that even regular topologies might run on N nodes,  
but not N+A nodes, giving the error messaeg that you post below.

Chris.

-- original message --

Hi all,

I am trying to use a simple scheme to restrain a few distances within my
protein, which is contained within an octahedral box of SPC water.

In place of distance restraints, I'm using bond type 6 to utilize a harmonic
potential to restrain the distance between a few atoms.  When I  
imposed one such
restraint in the topology, the simulation ran fine.  After finishing this
trajectory, I wanted to impose a second restraint and try again.  At  
this point,
the simulation fails with the following:

mpirun -np 24 mdrun_4.0.2_gcc_mpi -s nvt.tpr -deffnm nvt

<snip>

Initializing Domain Decomposition on 24 nodes
Dynamic load balancing: auto
Will sort the charge groups at every domain (re)decomposition
Initial maximum inter charge-group distances:
      two-body bonded interactions: 1.634 nm
    multi-body bonded interactions: 0.606 nm
Minimum cell size due to bonded interactions: 1.797 nm
Maximum distance for 5 constraints, at 120 deg. angles, all-trans: 0.876 nm
Estimated maximum distance required for P-LINCS: 0.876 nm
Guess for relative PME load: 0.45
Will use 12 particle-particle and 12 PME only nodes
This is a guess, check the performance at the end of the log file
Using 12 separate PME nodes
Scaling the initial minimum size with 1/0.8 (option -dds) = 1.25
Optimizing the DD grid for 12 cells with a minimum initial size of 2.246 nm
The maximum allowed number of cells is: X 2 Y 2 Z 2

-------------------------------------------------------
Program mdrun_4.0.2_gcc_mpi, VERSION 4.0.2
Source code file: domdec.c, line: 5842

Fatal error:
There is no domain decomposition for 12 nodes that is compatible with  
the given
box and a minimum cell size of 2.24633 nm
Change the number of nodes or mdrun option -rdd or -dds
Look in the log file for details on the domain decomposition
-------------------------------------------------------

"I'm Only Faking When I Get It Right" (Soundgarden)

I realize there are more optimizations I could impose to better balance the
PP/PME load.  I have tried those (altering my cutoff's and grid  
spacing) but it
makes no difference.  Plus, I'd like to remain consistent with the original
simulation that used only one restraint.  If I remove the second  
restraint, the
simulation runs just fine.  For reference, the restraints I'm imposing are:

    330  1016     6    0.669   5.0300e+06
   1016   559     6    0.376   5.0300e+06

The box vectors from the .gro file are:

     7.14132   6.73290   5.83086   0.00000   0.00000   2.38044   0.00000
-2.38044   3.36645

Is there anything obvious that I'm doing wrong?  Or should I just use real
distance restraints, according to manual section 4.3.4?  Thanks for  
any insights.

-Justin